| from onnxruntime import optimizer | |
| # Specify the models you want to optimize | |
| models = [ | |
| 'onnx/decoder_model.onnx', | |
| 'onnx/decoder_model_merged.onnx', | |
| 'onnx/decoder_model_merged_quantized.onnx', | |
| 'onnx/decoder_model_quantized.onnx', | |
| 'onnx/decoder_with_past_model.onnx', | |
| 'onnx/decoder_with_past_model_quantized.onnx', | |
| 'onnx/encoder_model.onnx', | |
| 'onnx/encoder_model_quantized.onnx' | |
| ] | |
| for model_path in models: | |
| # Load and optimize the model | |
| optimized_model = optimizer.optimize_model(model_path) | |
| # Save the optimized model | |
| optimized_model.save_model_to_file(model_path.replace('.onnx', '_optimized.onnx')) |